Raise error when inputs and mask don't have same shape in Softmax

keras/src/layers/activations/softmax.py

@@ -50,6 +50,12 @@ def __init__(self, axis=-1, **kwargs):
 
  def call(self, inputs, mask=None):
  if mask is not None:
+ if mask.shape != inputs.shape:
+ raise ValueError(
+ "`mask` and `inputs` must have same shape. "
+ f"Got inputs shape: {inputs.shape} and "
+ f"mask shape: {mask.shape}"
+ )
  adder = (
  1.0 - backend.cast(mask, inputs.dtype)
  ) * _large_negative_number(inputs.dtype)